The present invention relates to an inspection method used in a production process of an electronic device such as a semiconductor integrated circuit, and relates to an inspection apparatus or inspection tool and an inspection system, which are used for realizing the inspection method, and a manufacturing method of a semiconductor device.
In the production of an electronic device typified by a semiconductor integrated circuit, after a defect is detected by darkfield and brightfield inspection tools, the detected defect is often reviewed in order to analyze an individual detected defect using a defect review function provided in the inspection tool itself or a dedicated image acquiring tool, such as a review apparatus or review tool, having an electron microscope, etc.
It is to be noted that the darkfield inspection tool detects a particle adhered to a wafer whereas the brightfield inspection tool detects a particle and a pattern defect formed on the wafer. Hereinafter the particle and the pattern defect are generically referred to as defect.
As compared with the darkfield and brightfield inspection tools, the review tool picks up a position of an individual defect as a high-resolution image. Accordingly, the review tool does not pick up all defect positions detected by the inspection tool, but samples the defect positions on a surface of a wafer to limit the number of defect positions to a few before picking up an image.
Conventionally, random sampling was used for the above-mentioned sampling; more specifically, a defect was selected at random from among the detected defects.
In addition, Japanese Patent Laid-open No. Hei 10-214866 discloses a technology in which, if there is a cluster-like defect such as a flaw or closely formed defects, classifying defects detected by an inspection tool into defects inside the cluster-like defect and defects outside the cluster-like defect. Even in such a case, a few defects are sampled at random from among the defects inside the cluster-like defect; and likewise, a few defects are sampled at random from among the defects outside the cluster-like defect.
Although it was possible to grasp statistically a tendency of defects using the conventional random sampling, a necessary defect was not efficiently reviewed. For example, measures could not be taken by priority against a critical defect causing an electrical failure; with the result that it was difficult to improve a yield effectively.
Moreover, with the microminiaturization of a circuit pattern of a semiconductor, the size of a detectable defect, required for an inspection tool, becomes smaller. For this reason, performance of the inspection tool is being enhanced accordingly, leading to an increase in the number of defects to be detected. Therefore, establishment of an effective reviewing method is desired.